Social network Bluesky has sparked a heated debate with its recent proposal to give users more control over how their data is scraped for purposes like generative AI training and public archiving. The proposal, published on GitHub, outlines a new standard for users to indicate their preferences for data usage across four categories.

Bluesky's CEO, Jay Graber, discussed the proposal at South by Southwest, but it wasn't until she posted about it on Bluesky that it garnered widespread attention. Some users reacted with alarm, seeing the proposal as a reversal of Bluesky's previous stance on not selling user data to advertisers and not training AI on user posts. One user, Sketchette, exclaimed, "Oh, hell no! The beauty of this platform was the NOT sharing of information. Especially gen AI. Don't you cave now."

Graber responded by clarifying that generative AI companies are already scraping public data from across the web, including Bluesky, since everything on the platform is public like a website. She emphasized that Bluesky is trying to create a new standard to govern that scraping, similar to the robots.txt file that websites use to communicate their permissions to web crawlers. This new standard would provide a machine-readable format for users to express their preferences, which good actors are expected to abide by, although it would not be legally enforceable.

The proposal has sparked debate about AI training and copyright, with some highlighting the limitations of robots.txt, which is not legally enforceable. Molly White, who writes the Citation Needed newsletter and Web3 is Going Just Great blog, described the proposal as "a good proposal" and noted that it's "weird to see people flaming BlueSky for it," since it's not about "welcoming in AI scraping" but rather "trying to add a consent signal to allow users to communicate preferences for the scraping that is already happening."

Under the proposal, users of the Bluesky app or other apps that use the underlying ATProtocol could go into their settings and allow or disallow the usage of their Bluesky data across four categories: generative AI, protocol bridging, bulk datasets, and web archiving. If a user indicates that they don't want their data used to train generative AI, companies and research teams building AI training sets are expected to respect this intent when scraping websites or doing bulk transfers using the protocol itself.

While some see the proposal as a step towards giving users more control over their data, others are skeptical about its effectiveness. White noted that the proposal relies on scrapers to respect these signals out of a desire to be good actors, which may not always be the case. As she pointed out, some companies have already shown a willingness to blow past robots.txt or pirate material to scrape.

The proposal raises important questions about user privacy, data ownership, and the ethics of AI training. As the debate continues, it remains to be seen whether Bluesky's proposal will become a new standard for the industry or if it will face further scrutiny and criticism.

In the broader context, the proposal highlights the need for more nuanced discussions around AI training, copyright, and user privacy. As AI technology continues to advance, it's essential to establish clear guidelines and standards for data usage and scraping. Bluesky's proposal may be a step in the right direction, but it's only the beginning of a larger conversation that needs to take place.